Earlier this fall, leaders at the Eden Prairie, Minn.-based Optum corporation released the results from their third annual Optum Survey on AI in Health Care. Among the survey’s findings include that healthcare executives project a quick return on AI investments and nearly all said hiring AI talent is a top priority. The survey also found the top three applications health care executives plan to tap AI for – all of which can help improve health care in the current and post-COVID-19 world – include monitoring data from Internet of Things devices, such as wearable technology (40 percent); accelerating research for new therapeutic or clinical discoveries (37 percent); and assigning codes for accurate diagnosis and reimbursement (37 percent).
As a press release published to Optum’s website on Oct. 27 stated, “Health care executives today believe artificial intelligence (AI) will deliver value for the industry faster than previously thought, according to a new survey of 500 senior health care executives representing leading hospitals, health plans, life sciences and employers. The third annual Optum Survey on Artificial Intelligence (AI) in Health Care found that 59 percent of executives surveyed expect their organizations will see a full return on their AI investments in under three years, nearly double the 31 percent of leaders surveyed in 2018 who expected to break even that quickly. Further, health care executives’ confidence in AI’s ability to deliver operational efficiencies or improved clinical performance increases as their organizations progress on the maturity curve: among the respondents who identified as being in the late stages of AI deployment, 57% indicated they would reach their ROI in less than two years. Expectations for faster return on AI investments grew even among those in earlier stages of deployment, indicating growing confidence in the technology.”
The press release went on to note that, “In addition, leaders across industry segments expect to see tangible cost savings from their AI investments faster – an average of 3.6 years in this year’s survey, down from 4.7 years in 2019 and 5.3 years in 2018.” The press release quoted Robert Musslewhite, CEO of OptumInsight, who stated, “This year’s findings further confirm our belief in the potential of AI to deliver insights and operational efficiencies that unlock better performance across health care. It is encouraging to see so many organizations express confidence in AI’s ability to facilitate the pursuit of our industry’s shared goals: better health outcomes, better consumer experiences, and less physician burnout—all at a lower total cost of care.”
What’s more, the press release noted, “As artificial intelligence (AI) evolves into a must-have technology in almost every industry, health care organizations continue to develop— and even accelerate — their AI strategies in 2020: 83 percent have an AI strategy in place, and another 15 percent are planning on creating one, according to the third annual Optum Survey on AI in Health Care. In fact, 56 percent say they are accelerating or expanding their AI deployment timelines in response to the novel coronavirus (COVID-19) pandemic, demonstrating the importance of this business tool during the most stressful times. Furthermore, senior health care executives are increasingly optimistic that their AI investments will soon pay dividends, with 59 percent anticipating AI delivering tangible costs savings within three years — a 90 percent increase since 2018.”
“The need for skilled analytic talent in health care has never been greater,” said Stephen J. Griffiths, Ph.D., chief operating officer of Optum Enterprise Analytics. “The growing strategic importance of AI means that organizations need to ensure access to this skillset, either by growing their own analytic teams or seeking out experienced partners.”
And, with regard to the top three applications that healthcare executives plan to use AI for, the press release noted that “All three applications pose advantages in the current and post-COVID-19 world: AI can identify signals and trends from internet-connected remote patient monitoring devices, enabling more complete virtual health offerings; AI can help prioritize potential investigative targets for treatments or vaccines; and automating business processes can enable organizations to operate efficiently even with limited resources.”
The full report can be found here.
Shortly after the publication of the survey results, Griffiths, who has spent over 25 years in the healthcare industry, and who has a master’s degree in biostatistics and a doctorate in health services research, policy, and administration, spoke with Healthcare Innovation Editor-in-Chief Mark Hagland about the implications of the report’s findings, for patient are organizations, in the near future. Below are excerpts from that interview.
What did you think of the finding that 59 percent of executives surveyed expect that their organizations will see a full return on their AI investments in under three years, nearly double the 31 percent of leaders surveyed in 2018 who expected to break even that quickly?
I think it really speaks to this maturity group concept. Healthcare, as you know, is a later adopter of some of these capabilities, compared to the retailing or financial services industry. So there’s a sort of maturity curve, and it’s becoming real in the industry. Early on, there’s a sort of hype cycle; but organizations have to make investments, align resources, begin on the path, and fail fast, and so on. But it’s becoming real now for organizations in a number of ways.
Where are we starting to see some of the first practical breakthroughs?
Around disease prediction or the interception of disease, so we’re finding things faster, intervening faster, and helping people faster to get the things they need. There are some administrative processes in healthcare, where we’re seeing work around NLP [natural language processing], and interpreting unstructured text, such as in physician notes. That creates new data models and also greater efficiency, removing some manual work as well.
What do you see as some of the biggest challenges facing patient care organization leaders in this area, in the next few years?
To me, this isn’t just about having more data scientists or data engineers to do the work; it’s really about corporate transformation and evolution. You have not just people who can build models, but business leaders who can begin to understand the art of the possible, and to connect this to business strategy. And that has been a barrier. Many thought they could just hire more data scientists. But you could be creating science experiments that really don’t lead to benefit.
There’s been an ongoing debate over whether organizations should be bringing data scientists into healthcare from outside the industry, or conversely, trying to train healthcare professionals to become data scientists. What do you think?
Both of those things are needed. In addition to a growing awareness of business leaders—you need people who understand healthcare. The people who understand actuarial methods and statistical methods, haven’t learned some of the new AI methods. So we’ve been training them. And we’ve also been investing in people external to healthcare, and we’ve been combining them. You need applications and methods understanding, and on top of that, you need strong business leaders.
With regard to the top three applications cited—monitoring data from IoT devices; accelerating research for new therapeutic or clinical discoveries; and assigning codes for accurate diagnosis and reimbursement—what did you think of those results?
Those are very different things. And that’s helping to illustrate that people are seeing not only that AI is real, and helpful, and additive, but that it’s also multi-dimensional. And particularly now with people being at home and are self-monitoring, the wearables result makes sense. And now, we can use tools to read the entire patient chart, and move away from some manual processes.
What has the discovery process around AI looked like in the past two years?
I always talk about the slope of the curve, which is moving up. People become curious to see whether it’s real. And they become aware of it and interested. And internally here, AI automation-leveraging programs have been led internally, and it’s all been maturing. And things have really been accelerating in the past 18 months.
Is anyone making any big mistakes?
“Mistake” is a loaded word; but that’s part of failing fast. If people are making any mistakes, it’s that often, in organizations, if people don’t view this as a sport—because we need multidimensional skills coming together, with lots of fingerprints. It’s a corporate transformation. If people see this as just an engineering or analytics thing, you’re really missing an opportunity to drive acceleration.
Per some of our past reporting, do you see clinical use cases as presenting themselves at a certain level of specificity? For example, radiological study prioritization?
There’s a broadening of AI-assisted clinical decision-making that’s emerging. And we can use models to help certain processes go fast. Certain things need to be reviewed and to have clinical judgment, so you can use deep-learning methods that can scan and read entire documents, to help physicians go faster, to make more holistic decisions, and also to help them think more consistently—to make better, faster, more consistent decisions that lead to good outcomes.
How do you see the next two years playing out?
We’re still early on, so the acceleration is going to continue over the next couple of years. I think that the talent issue is going to become increasingly important. There will be a shortage of talent and competition for those people, not just the data scientists and data engineers; but there will be an increasing demand for the business leaders and the clinicians who can help lead these initiatives. We’re seeing universities offering business analytics courses, and similar, because of the need.
And there will be a shift to surveillance. Many organizations living through COVID see that as a reason to expand. And a lot of healthcare is retrospective. But we learned that interoperability of data is a challenge, and we learned in the pandemic that there are big gaps in the supply chain information system. And so we’ll need in advance to be able to identify hot spots of all kinds, not just for potential future pandemics, but also for many other applications in healthcare.
Is there anything else that you’d like to add?
Personally, as sort of a healthcare nerd, it’s a fun time to be in healthcare. This stuff is so cool. And we’ve all been waiting for analytics to become real in healthcare. And to be able to think multidimensionally about data, around the coming together of broad, different types of healthcare professionals. This has been a dream for so many of us in analytics in healthcare, for some time now. And everything we’re seeing in AI is helping to foster transformation. And in the end, this is all about improving people’s lives and achieving the Quadruple Aim. And there are areas like physician burnout, where this will be so important. And within Optum, we have a wide variety of different types of experts and are well-positioned to do this work. And so it’s an optimistic time.
